The Research On The Modeling Method And Communication Mode In The Intelligentization Of Machine Group

The intelligentization of machine group as a new research field is developedwith the maturity of the intelligentization of single machine. The latter laysparticular emphasis on raising the level of mechanical intelligence and researchingthe efficiency and flexibility of the machine. The intelligentization of singlemachine can't already satisfy the extensive high-speed automated production andcan't meet the convenience offered by the high-speed information network, and ithas already become the bottleneck which restricts the development of productionautomation.The research on the intelligentiation of machine group can resolve theproblem above. It emphasizes the thinking of "group", that is studying theorganization of the group, the relation, communication and behavior betweenmembers of the group, so some scholars regard the "group" as "society" similar tohuman society, basing on organizational behaviorism. They bring the ideology oforganizational behaviorism to the study of the intelligentiation of machine groupand form the analytical method of the intelligentization of machine group based onorganizational behaviorism. Besides, the study of the intelligentization of singlemachine shows that every unit of the system is an "Agent" logically. In a broadsense, the agent is the generic term of all the elements which are predictive, socialand automatic, and persons are also a kind of agent;in a narrow sense, it is a set ofsoftware and hardware which has the qualities above. Basing on the agent thought,some researchers convert the study of the intelligentization of machine group intothe study of Multi-Agent System (MAS), including the study of the structure,communication and coordinated mechanism of the MAS.The important issue of MAS is the way how to describe and create the MAS,and the modeling method of MAS is based on summing up these two aspects. Inbrief, the modeling is the way to simplify purposively. Therefore modeling turns tothe following aspects: (1) what to simplify, (2) why to simplify and (3) how tosimplify. The first two questions are simple. The process of answering the firstquestion is called requirement analysis in the field of computer and the answer tothe second question is in order to resolve the problem by calculation method. Thethird one is the key, relating to the extent and the method of simplification.The domestic and foreign scholars combine the theory of organizationalbehaviorism and the theory of MAS based on the thought of software, and build"the modeling method of MAS based on the role"(MmMASR). The method stemsfrom the Gaia which is put forward by Wooldridge, Jennings and Kinny. Theshortage of the Gaia is that the relation between agents and the ability of the agentcannot change during the process of operation, which restricts the ability of theagent and can't reflect the objective reality. Later, the improvers advance that eachagent has different "social" relation and behavioral abilities, which is acting asdifferent roles, during the different stages of operation. They consider that eachagent acts as a certain role in the society and changes when the circumstance andtime changes.The course of MmMASR is as the following: (1)Carry on systematicrequirement analysis, emphasize resolving goals, try to resolve the whole goal intosimple sub-goals which can be realized by the role, not too complicated, and thendescribe the sub-goals with the layer graph of goals;(2)Realize each goal by theroles which are stable and definite, finish the design of the graph of roles, create theview of role organization, behavioral state and intercommunion;(3)Create the"Agent" class, sum up the similar functions of different roles, create, reduce andcombine the "Agent" class according to the functions, and create the graph of roletransition;(4)Assemble the "Agent".We model the project which needs to be solved according to the method above,during the process, we find the following questions: in MmMASR, the first step isto resolve the goal into sub-goals realized by the role. These goals cannot changewhen the system is operated, that is to say, the goals are still in static state and theyare not dynamic factors, and the situation in the objective world will be often alittle more complicated. Then, we set the index of the goal (quantity) as systematicvariable and add the role called the dynamic goal chooser while designing in orderto adapt to the quantitative change of the goal. Finally, we use Java, syncretize thetechnology of CSCW(Computer Supporting Co-operation Work) and partiallyrealize the model (mainly the models of road and vehicle) created above, which hasthe certain visual effect and the property of the agent (automatic and social).The foundation for realizing the MAS is the communication. Thecharacteristic differing from the traditional dispersion control system is that eachagent is able to independently comprehend with other agents. The "comprehension"means that they can exchange information without predefined communicationlanguage or structure. As long as any two agents in the system can converse incommon language, the communion is useful(reflect the segregation of the design ofthe Agent and the MAS).In order to provide the bottom foundation for realizing thecommunication, we define the physics and software base of the communicationlink: requiring certain physics link and TCP/IP protocol, according to this, wedefine a kind of language for enquiry based on KQML(mainly concentrating on thecontent layer of KQML).The language provides the ability to enquire theinformation of every model in our software system, for example, inquiry of theinformation of road and vehicle.Our analysis, design, conception and realization are all along the thought ofMAS modeling based on the organizational behaviorism and try to studytheoretically at the support of the system dynamic goals. However, we think thatthe method that we introduce the thought of the modeling method of MAS basedon role into the design of intelligentization of machine group is meaningful. Furthermore, we use the method to partially realize the function, proving that the methodis correct and feasible.